@jarvis-agent/core 0.1.5 → 0.1.7

package/dist/index.esm.js

@@ -1,4 +1,4 @@
-const config$1 = {
+const defaultConfig = {
     name: "Eko",
     platform: "mac",
     maxReactNum: 500,
@@ -14,7 +14,18 @@ const config$1 = {
     parallelToolCalls: true,
     expertMode: false,
     expertModeTodoLoopNum: 10,
+    streamFirstTimeout: 30000,
+    streamTokenTimeout: 180000,
 };
+let config$1 = { ...defaultConfig };
+function mergeGlobalConfig(userConfig) {
+    if (userConfig) {
+        config$1 = { ...defaultConfig, ...userConfig };
+    }
+    else {
+        config$1 = { ...defaultConfig };
+    }
+}
 
 var LogLevel;
 (function (LogLevel) {
@@ -23926,6 +23937,7 @@ class Context {
         this.chain = chain;
         this.variables = new Map();
         this.controller = new AbortController();
+        mergeGlobalConfig(config.globalConfig);
     }
     async checkAborted(noCheckPause) {
         if (this.controller.signal.aborted) {
@@ -29806,8 +29818,7 @@ async function compressAgentMessages(agentContext, messages, tools) {
 }
 async function doCompressAgentMessages(agentContext, messages, tools) {
     const ekoConfig = agentContext.context.config;
-    const rlm = new RetryLanguageModel(ekoConfig.llms, ekoConfig.compressLlms);
-    rlm.setContext(agentContext);
+    const rlm = new RetryLanguageModel(ekoConfig.llms, ekoConfig.compressLlms, ekoConfig.globalConfig?.streamFirstTimeout, ekoConfig.globalConfig?.streamTokenTimeout, agentContext);
     // extract used tool
     const usedTools = extractUsedTool(messages, tools);
     const snapshotTool = new TaskSnapshotTool();
@@ -31225,8 +31236,7 @@ class Planner {
     }
     async doPlan(taskPrompt, messages, saveHistory, retryNum = 0) {
         const config = this.context.config;
-        const rlm = new RetryLanguageModel(config.llms, config.planLlms);
-        rlm.setContext(this.context);
+        const rlm = new RetryLanguageModel(config.llms, config.planLlms, config.globalConfig?.streamFirstTimeout, config.globalConfig?.streamTokenTimeout, this.context);
         const request = {
             maxTokens: 8192,
             temperature: 0.7,
@@ -31683,8 +31693,7 @@ async function checkTaskReplan(agentContext) {
         if (!chain.planRequest || !chain.planResult) {
             return false;
         }
-        const rlm = new RetryLanguageModel(context.config.llms, context.config.planLlms);
-        rlm.setContext(agentContext);
+        const rlm = new RetryLanguageModel(context.config.llms, context.config.planLlms, context.config.globalConfig?.streamFirstTimeout, context.config.globalConfig?.streamTokenTimeout, agentContext);
         const agentExecution = getAgentExecutionPrompt(agentContext);
         const prompt = `# Task Execution Status
 ${agentExecution}
@@ -32719,7 +32728,7 @@ class EkoDialogue {
             params.messageId = params.messageId ?? this.memory.genMessageId();
             await this.addUserMessage(params.user, params.messageId);
         }
-        const rlm = new RetryLanguageModel(this.config.llms, this.config.chatLlms);
+        const rlm = new RetryLanguageModel(this.config.llms, this.config.chatLlms, this.config.globalConfig?.streamFirstTimeout, this.config.globalConfig?.streamTokenTimeout);
         for (let i = 0; i < 15; i++) {
             const messages = this.memory.buildMessages();
             const chatTools = [...this.buildInnerTools(params), ...this.tools];
@@ -33485,8 +33494,7 @@ request_help: Request assistance from the user; for instance, when an operation
         }
         try {
             let imageResult = (await screenshot.call(agentContext.agent, agentContext));
-            let rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms);
-            rlm.setContext(agentContext);
+            let rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms, agentContext.context.config.globalConfig?.streamFirstTimeout, agentContext.context.config.globalConfig?.streamTokenTimeout, agentContext);
             let image = toImage(imageResult.imageBase64);
             let request = {
                 messages: [
@@ -33734,8 +33742,7 @@ class WatchTriggerTool {
         const start = new Date().getTime();
         const timeout = (args.timeout || 5) * 60000;
         const frequency = Math.max(500, (args.frequency || 1) * 1000);
-        const rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms);
-        rlm.setContext(agentContext);
+        const rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms, agentContext.context.config.globalConfig?.streamFirstTimeout, agentContext.context.config.globalConfig?.streamTokenTimeout, agentContext);
         while (new Date().getTime() - start < timeout) {
             await agentContext.context.checkAborted();
             await new Promise((resolve) => setTimeout(resolve, frequency));
@@ -34226,25 +34233,31 @@ class Agent {
         this.description = params.description;
         this.tools = params.tools;
         this.llms = params.llms;
-        this.mcpClient = params.mcpClient;
+        this.mcpClients = params.mcpClients || (params.mcpClient ? [params.mcpClient] : []);
         this.planDescription = params.planDescription;
         this.requestHandler = params.requestHandler;
     }
     async run(context, agentChain) {
-        const mcpClient = this.mcpClient || context.config.defaultMcpClient;
+        const mcpClients = this.mcpClients.length > 0
+            ? this.mcpClients
+            : (context.config.defaultMcpClient ? [context.config.defaultMcpClient] : []);
         const agentContext = new AgentContext(context, this, agentChain);
         try {
             this.agentContext = agentContext;
-            mcpClient &&
-                !mcpClient.isConnected() &&
-                (await mcpClient.connect(context.controller.signal));
-            return await this.runWithContext(agentContext, mcpClient, config$1.maxReactNum);
+            for (const client of mcpClients) {
+                if (!client.isConnected()) {
+                    await client.connect(context.controller.signal);
+                }
+            }
+            return await this.runWithContext(agentContext, mcpClients, config$1.maxReactNum);
         }
         finally {
-            mcpClient && (await mcpClient.close());
+            for (const client of mcpClients) {
+                await client.close();
+            }
         }
     }
-    async runWithContext(agentContext, mcpClient, maxReactNum = 100, historyMessages = []) {
+    async runWithContext(agentContext, mcpClients, maxReactNum = 100, historyMessages = []) {
         let loopNum = 0;
         let checkNum = 0;
         this.agentContext = agentContext;
@@ -34267,18 +34280,20 @@ class Agent {
             },
         ];
         agentContext.messages = messages;
-        const rlm = new RetryLanguageModel(context.config.llms, this.llms);
-        rlm.setContext(agentContext);
+        const rlm = new RetryLanguageModel(context.config.llms, this.llms, context.config.globalConfig?.streamFirstTimeout, context.config.globalConfig?.streamTokenTimeout, agentContext);
+        const resolvedMcpClients = Array.isArray(mcpClients)
+            ? mcpClients
+            : (mcpClients ? [mcpClients] : []);
         let agentTools = tools;
         while (loopNum < maxReactNum) {
             await context.checkAborted();
-            if (mcpClient) {
+            if (resolvedMcpClients.length > 0) {
                 const controlMcp = await this.controlMcpTools(agentContext, messages, loopNum);
                 if (controlMcp.mcpTools) {
-                    const mcpTools = await this.listTools(context, mcpClient, agentNode, controlMcp.mcpParams);
+                    const mcpTools = await this.listMcpTools(context, resolvedMcpClients, agentNode, controlMcp.mcpParams);
                     const usedTools = extractUsedTool(messages, agentTools);
-                    const _agentTools = mergeTools(tools, usedTools);
-                    agentTools = mergeTools(_agentTools, mcpTools);
+                    const mergedTools = mergeTools(tools, usedTools);
+                    agentTools = mergeTools(mergedTools, mcpTools);
                 }
             }
             await this.handleMessages(agentContext, messages, tools);
@@ -34435,33 +34450,32 @@ class Agent {
     async extSysPrompt(agentContext, tools) {
         return "";
     }
-    async listTools(context, mcpClient, agentNode, mcpParams) {
-        try {
-            if (!mcpClient.isConnected()) {
-                await mcpClient.connect(context.controller.signal);
+    async listMcpTools(context, clients, agentNode, mcpParams) {
+        const allTools = [];
+        for (const client of clients) {
+            try {
+                if (!client.isConnected()) {
+                    await client.connect(context.controller.signal);
+                }
+                const list = await client.listTools({
+                    taskId: context.taskId,
+                    nodeId: agentNode?.id,
+                    environment: config$1.platform,
+                    agent_name: agentNode?.name || this.name,
+                    params: {},
+                    prompt: agentNode?.task || context.chain.taskPrompt,
+                    ...(mcpParams || {}),
+                }, context.controller.signal);
+                for (const toolSchema of list) {
+                    const execute = this.toolExecuter(client, toolSchema.name);
+                    allTools.push(new McpTool(new ToolWrapper(toolSchema, execute)));
+                }
+            }
+            catch (e) {
+                Log.error("Mcp listTools error", e);
             }
-            let list = await mcpClient.listTools({
-                taskId: context.taskId,
-                nodeId: agentNode?.id,
-                environment: config$1.platform,
-                agent_name: agentNode?.name || this.name,
-                params: {},
-                prompt: agentNode?.task || context.chain.taskPrompt,
-                ...(mcpParams || {}),
-            }, context.controller.signal);
-            let mcpTools = [];
-            for (let i = 0; i < list.length; i++) {
-                let toolSchema = list[i];
-                let execute = this.toolExecuter(mcpClient, toolSchema.name);
-                let toolWrapper = new ToolWrapper(toolSchema, execute);
-                mcpTools.push(new McpTool(toolWrapper));
-            }
-            return mcpTools;
-        }
-        catch (e) {
-            Log.error("Mcp listTools error", e);
-            return [];
         }
+        return allTools;
     }
     async controlMcpTools(agentContext, messages, loopNum) {
         return {
@@ -34504,9 +34518,9 @@ class Agent {
         };
     }
     async loadTools(context) {
-        if (this.mcpClient) {
-            let mcpTools = await this.listTools(context, this.mcpClient);
-            if (mcpTools && mcpTools.length > 0) {
+        if (this.mcpClients.length > 0) {
+            const mcpTools = await this.listMcpTools(context, this.mcpClients);
+            if (mcpTools.length > 0) {
                 return mergeTools(this.tools, mcpTools);
             }
         }
@@ -34538,8 +34552,11 @@ class Agent {
     get PlanDescription() {
         return this.planDescription;
     }
+    get McpClients() {
+        return this.mcpClients;
+    }
     get McpClient() {
-        return this.mcpClient;
+        return this.mcpClients[0];
     }
     get AgentContext() {
         return this.agentContext;
@@ -34548,8 +34565,8 @@ class Agent {
 
 const AGENT_NAME$3 = "File";
 class BaseFileAgent extends Agent {
-    constructor(work_path, llms, ext_tools, mcpClient, planDescription) {
-        const _tools_ = [];
+    constructor(work_path, llms, ext_tools, mcpClients, planDescription) {
+        const initTools = [];
         const prompt = work_path
             ? `Your working directory is: ${work_path}
 - When viewing file lists and outputting file paths, always include the working directory
@@ -34562,9 +34579,9 @@ class BaseFileAgent extends Agent {
         super({
             name: AGENT_NAME$3,
             description: `You are a file agent, handling file-related tasks such as creating, finding, reading, modifying files, etc.${prompt}`,
-            tools: _tools_,
+            tools: initTools,
             llms: llms,
-            mcpClient: mcpClient,
+            mcpClients: Array.isArray(mcpClients) ? mcpClients : (mcpClients ? [mcpClients] : []),
             planDescription: planDescription ||
                 `File operation agent, handles file-related tasks such as creating, finding, reading, modifying files, etc. Only supports text file output
 - Output file names must be in English
@@ -34572,11 +34589,11 @@ class BaseFileAgent extends Agent {
 - For data-related content, combine with visualization tools for display
 - For visualizations, generate charts first before page generation to minimize repetitive work`,
         });
-        let init_tools = this.buildInitTools();
+        let builtTools = this.buildInitTools();
         if (ext_tools && ext_tools.length > 0) {
-            init_tools = mergeTools(init_tools, ext_tools);
+            builtTools = mergeTools(builtTools, ext_tools);
         }
-        init_tools.forEach((tool) => _tools_.push(tool));
+        builtTools.forEach((tool) => initTools.push(tool));
     }
     async do_file_read(agentContext, path, write_variable) {
         let file_context = await this.file_read(agentContext, path);
@@ -34731,24 +34748,24 @@ class BaseFileAgent extends Agent {
 
 const AGENT_NAME$2 = "Shell";
 class BaseShellAgent extends Agent {
-    constructor(llms, ext_tools, mcpClient, planDescription) {
-        const _tools_ = [];
+    constructor(llms, ext_tools, mcpClients, planDescription) {
+        const initTools = [];
         super({
             name: AGENT_NAME$2,
             description: `Run commands in a bash shell,
 * You must first call create_session to create a new session when using it for the first time.
 * Please execute delete commands with caution, and never perform dangerous operations like \`rm -rf /\`.
 * Please avoid commands that may produce a very large amount of output.`,
-            tools: _tools_,
+            tools: initTools,
             llms: llms,
-            mcpClient: mcpClient,
+            mcpClients: Array.isArray(mcpClients) ? mcpClients : (mcpClients ? [mcpClients] : []),
             planDescription: planDescription || "Shell command agent, use to execute shell commands.",
         });
-        let init_tools = this.buildInitTools();
+        let builtTools = this.buildInitTools();
         if (ext_tools && ext_tools.length > 0) {
-            init_tools = mergeTools(init_tools, ext_tools);
+            builtTools = mergeTools(builtTools, ext_tools);
         }
-        init_tools.forEach((tool) => _tools_.push(tool));
+        builtTools.forEach((tool) => initTools.push(tool));
     }
     buildInitTools() {
         return [
@@ -34813,8 +34830,8 @@ class BaseShellAgent extends Agent {
 
 const AGENT_NAME$1 = "Computer";
 class BaseComputerAgent extends Agent {
-    constructor(llms, ext_tools, mcpClient, keyboardKeys) {
-        const _tools_ = [];
+    constructor(llms, ext_tools, mcpClients, keyboardKeys) {
+        const initTools = [];
         super({
             name: AGENT_NAME$1,
             description: `You are a computer operation agent, who interacts with the computer using mouse and keyboard, completing specified tasks step by step based on the given tasks and screenshots. After each of your operations, you will receive the latest computer screenshot to evaluate the task execution status.
@@ -34822,9 +34839,9 @@ This is a computer GUI interface, observe the execution through screenshots, and
 * COMPUTER OPERATIONS:
   - You can operate the application using shortcuts.
   - If stuck, try alternative approaches`,
-            tools: _tools_,
+            tools: initTools,
             llms: llms,
-            mcpClient: mcpClient,
+            mcpClients: Array.isArray(mcpClients) ? mcpClients : (mcpClients ? [mcpClients] : []),
             planDescription: "Computer operation agent, interact with the computer using the mouse and keyboard."
         });
         if (!keyboardKeys) {
@@ -34855,11 +34872,11 @@ This is a computer GUI interface, observe the execution through screenshots, and
                 ];
             }
         }
-        let init_tools = this.buildInitTools(keyboardKeys);
+        let builtTools = this.buildInitTools(keyboardKeys);
         if (ext_tools && ext_tools.length > 0) {
-            init_tools = mergeTools(init_tools, ext_tools);
+            builtTools = mergeTools(builtTools, ext_tools);
         }
-        init_tools.forEach((tool) => _tools_.push(tool));
+        builtTools.forEach((tool) => initTools.push(tool));
     }
     buildInitTools(keyboardKeys) {
         return [
@@ -35986,7 +36003,7 @@ function run_build_dom_tree() {
 }
 
 class BaseBrowserLabelsAgent extends BaseBrowserAgent {
-    constructor(llms, ext_tools, mcpClient) {
+    constructor(llms, ext_tools, mcpClients) {
         let description = `You are a browser operation agent, use structured commands to interact with the browser.
 * This is a browser GUI interface where you need to analyze webpages by taking screenshot and page element structures, and specify action sequences to complete designated tasks.
 * For your first visit, please start by calling either the \`navigate_to\` or \`current_page\` tool. After each action you perform, I will provide you with updated information about the current state, including page screenshots and structured element data that has been specially processed for easier analysis.
@@ -36021,20 +36038,20 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
    - When filling out a form, fields that are not dependent on each other should be filled simultaneously
    - Avoid parallel processing for dependent operations, such as those that need to wait for page loading, DOM changes, redirects, subsequent operations that depend on the results of previous operations, or operations that may interfere with each other and affect the same page elements. In these cases, please do not use parallelization.`;
         }
-        const _tools_ = [];
+        const initTools = [];
         super({
             name: AGENT_NAME,
             description: description,
-            tools: _tools_,
+            tools: initTools,
             llms: llms,
-            mcpClient: mcpClient,
+            mcpClients: Array.isArray(mcpClients) ? mcpClients : (mcpClients ? [mcpClients] : []),
             planDescription: "Browser operation agent, interact with the browser using the mouse and keyboard.",
         });
-        let init_tools = this.buildInitTools();
+        let builtTools = this.buildInitTools();
         if (ext_tools && ext_tools.length > 0) {
-            init_tools = mergeTools(init_tools, ext_tools);
+            builtTools = mergeTools(builtTools, ext_tools);
         }
-        init_tools.forEach((tool) => _tools_.push(tool));
+        builtTools.forEach((tool) => initTools.push(tool));
     }
     async input_text(agentContext, index, text, enter) {
         await this.execute_script(agentContext, typing, [{ index, text, enter }]);
@@ -36739,7 +36756,7 @@ function scroll_by(params) {
 }
 
 class BaseBrowserScreenAgent extends BaseBrowserAgent {
-    constructor(llms, ext_tools, mcpClient) {
+    constructor(llms, ext_tools, mcpClients) {
         const description = `You are a browser operation agent, use a mouse and keyboard to interact with a browser.
 * This is a browser GUI interface, observe the webpage execution through screenshots, and specify action sequences to complete designated tasks.
 * For the first visit, please call the \`navigate_to\` or \`current_page\` tool first. After that, each of your actions will return a screenshot of the page.
@@ -36751,20 +36768,20 @@ class BaseBrowserScreenAgent extends BaseBrowserAgent {
   - Wait for elements to load
   - Scroll pages and handle infinite scroll
   - YOU CAN DO ANYTHING ON THE BROWSER - including clicking on elements, filling forms, submitting data, etc.`;
-        const _tools_ = [];
+        const initTools = [];
         super({
             name: AGENT_NAME,
             description: description,
-            tools: _tools_,
+            tools: initTools,
             llms: llms,
-            mcpClient: mcpClient,
+            mcpClients: Array.isArray(mcpClients) ? mcpClients : (mcpClients ? [mcpClients] : []),
             planDescription: "Browser operation agent, interact with the browser using the mouse and keyboard.",
         });
-        let init_tools = this.buildInitTools();
+        let builtTools = this.buildInitTools();
         if (ext_tools && ext_tools.length > 0) {
-            init_tools = mergeTools(init_tools, ext_tools);
+            builtTools = mergeTools(builtTools, ext_tools);
         }
-        init_tools.forEach((tool) => _tools_.push(tool));
+        builtTools.forEach((tool) => initTools.push(tool));
     }
     buildInitTools() {
         return [